Parameterization of reactive force field for thiol oxidation and disulfide linking

This study employed a simulated annealing algorithm to develop new reactive force field (ReaxFF) parameters specifically tailored for thiol oxidation and disulfide cross-linking reactions. The optimization strategy demonstrably enhanced the capability of ReaxFF to model these chemical processes. To mitigate potential overfitting during parameterization, we strategically expanded the training set's conformational space relevant to the reactions through normal mode analysis. This approach effectively addressed concerns about overfitting, and the resulting parameters exhibit a degree of scalability for similar reactions. The optimized parameters achieve good agreement with experimental reaction energies. While surpassing the original parameters in accuracy, the characterization of reaction energy barriers is not ideal due to the exclusion of parameters related to O−H bond during optimization and inherent limitations within the ReaxFF framework. However, the current level of accuracy is sufficient for preliminary semi-quantitative studies. Notably, molecular dynamics simulations utilizing the new parameters successfully captured the binding mode of reactants in solvent- mediated proton transfer reactions, even for systems beyond the training set. These advancements in the ReaxFF force field offer a more efficient and reliable computational tool for exploring thiol oxidation and disulfide cross-linking reactions relevant to biological systems.